MATLAB Linear AlgebraLinear Algebra and Numerical Matrix Theory Topics: Illustrate with MATLAB as...

Ralph C. SmithDepartment of Mathematics

North Carolina State University

Introduction to MATLAB and Linear Algebra

Overview: 1. MATLAB examples throughout tutorial

2. Elements of linear algebra

• Fundamental properties of vectors and matrices

• Eigenvalues, eigenvectors and singular values

Linear Algebra and Numerical Matrix Theory:

• Vector properties including orthogonality

• Matrix analysis, inversion and solving Ax = b for very large systems

• Eigen- and singular value decompositions

Linear Algebra and Numerical Matrix Theory

Topics: Illustrate with MATLAB as topics are introduced

• Basic concepts

• Linear transformations

• Linear independence, basis vectors, and span of a vector space

• Fundamental Theorem of Linear Algebra

• Determinants and matrix rank

• Eigenvalues and eigenvectors

• Solving Ax = b and condition numbers

• Singular value decomposition (SVD)

• Cholesky and QR decompositions

Linear Algebra has become as basic and as applicable as calculus, and fortunately it is easier.

--Gilbert Strang, MIT

Vectors and Matrices

• Vector in Rn is an ordered set of n real numbers.– e.g. v = (1,6,3,4) is in R4

– “(1,6,3,4)” is a column vector:

– as opposed to a row vector:

• m-by-n matrix is an object with m rows and n columns, each entry fill with a real number:

÷÷÷÷÷

ø

ö

ççççç

è

æ

4361

( )4361

÷÷÷

ø

ö

ççç

è

æ

2396784821

Note:

Vector Norms

Vector Norms: The norm ||x|| quantifies the ``length’’

Common Norms:

Euclidean: Vector length

Vector and Matrix Products

Vector products:– Dot product: ( ) 2211

2

121 vuvuvv

uuvuvu T +=÷÷ø

öççè

æ==•

( ) ÷÷ø

öççè

æ=÷÷

ø

öççè

æ=

2212

211121

2

1

vuvuvuvu

vvuu

uvT

Matrix Product:

÷÷ø

öççè

æ++++

=

÷÷ø

öççè

æ=÷÷

ø

öççè

æ=

2222122121221121

2212121121121111

2221

1211

2221

1211 ,

babababababababa

AB

bbbb

Baaaa

A

– Outer product:

Note: )cos(qBABA =×

If u•v=0, ||u||2 != 0, ||v||2 != 0 à u and v are orthogonal

If u•v=0, ||u||2 = 1, ||v||2 = 1 à u and v are orthonormal

Special Matrices

÷÷÷

ø

ö

ççç

è

æ

fedcba

000

÷÷÷

ø

ö

ççç

è

æ

cb

a

000000

÷÷÷÷÷

ø

ö

ççççç

è

æ

jihgf

edcba

000

000

÷÷÷

ø

ö

ççç

è

æ

fedcb

a000

diagonal upper-triangular

tri-diagonal lower-triangular

÷÷÷

ø

ö

ççç

è

æ

100010001

I (identity matrix)

Special Matrices

• Matrix A is symmetric if A = AT

• A is positive definite if xTAx>0 for all non-zero x (positive semi-definite if inequality is not strict)

( ) 222

100010001

cbacba

cba ++=÷÷÷

ø

ö

ççç

è

æ

÷÷÷

ø

ö

ççç

è

æ( ) 222

100010001

cbacba

cba +-=÷÷÷

ø

ö

ççç

è

æ

÷÷÷

ø

ö

ççç

è

æ-

• Useful fact: Any matrix of form ATA is positive semi-definite.To see this, xT(ATA)x = (xTAT)(Ax) = (Ax)T(Ax) ≥ 0

Recall: Covariance matrix

V = �2(X T X )-1

Matrices as Linear Transformations

÷÷ø

öççè

æ=÷÷

ø

öççè

æ÷÷ø

öççè

æ55

11

5005

(stretching)

÷÷ø

öççè

æ-=÷÷

ø

öççè

æ÷÷ø

öççè

æ -11

11

0110

(rotation)

÷÷ø

öççè

æ=÷÷

ø

öççè

æ÷÷ø

öççè

æ10

01

0110

(reflection)

÷÷ø

öççè

æ=÷÷

ø

öççè

æ÷÷ø

öççè

æ01

11

0001

(projection)

Linear Independence

÷÷÷

ø

ö

ççç

è

æ=

÷÷÷

ø

ö

ççç

è

æ

÷÷÷

ø

ö

ççç

è

æ

000

|||

|||

3

2

1

321

ccc

vvv

(u,v)=(0,0), i.e. the columns are linearly independent.

• A set of vectors is linearly independent if none of them can be written as a linear combination of the others.

• Vectors v1,…,vk are linearly independent if c1v1+…+ckvk= 0 implies c1=…=ck=0

÷÷÷

ø

ö

ççç

è

æ=÷÷

ø

öççè

æ

÷÷÷

ø

ö

ççç

è

æ

000

313201

vu

e.g.

x3 = −2x1 + x2

Basis Vectors

• If all vectors in a vector space may be expressed as linear combinations of a set of vectors v1,…,vk, then v1,…,vk spansthe space.

• The cardinality of this set is the dimension of the vector space.

• A basis is a maximal set of linearly independent vectors and a minimal set of spanning vectors of a vector space

÷÷÷

ø

ö

ççç

è

æ+

÷÷÷

ø

ö

ççç

è

æ+

÷÷÷

ø

ö

ççç

è

æ=

÷÷÷

ø

ö

ççç

è

æ

100

2010

2001

2222

(0,0,1)

(0,1,0)

(1,0,0)

e.g.

Basis Vectors

• An orthonormal basis consists of orthogonal vectors of unit length.Note:

÷÷÷

ø

ö

ççç

è

æ+

÷÷÷

ø

ö

ççç

è

æ+

÷÷÷

ø

ö

ççç

è

æ=

÷÷÷

ø

ö

ççç

è

æ

100

2010

2001

2222

(0,0,1)

(0,1,0)

(1,0,0)

(.1,.2,1)

(.3,1,0)(.9,.2,0)

÷÷÷

ø

ö

ççç

è

æ+

÷÷÷

ø

ö

ççç

è

æ+

÷÷÷

ø

ö

ççç

è

æ=

÷÷÷

ø

ö

ççç

è

æ

12.1.

2013.

29.102.9.

57.1222

Rank of a Matrix

• The rank of A is the dimension of the column space of A.• It also equals the dimension of the row space of A (the subspace of vectors

which may be written as linear combinations of the rows of A).

÷÷÷

ø

ö

ççç

è

æ

313201 (1,3) = (2,3) – (1,0)

Only 2 linearly independent rows, so rank = 2.

Fundamental Theorem of Linear Algebra:If A is m x n with rank r,

Column space(A) has dimension rNullspace(A) has dimension n-r (= nullity of A)Row space(A) = Column space(AT) has dimension rLeft nullspace(A) = Nullspace(AT) has dimension m - r

Rank-Nullity Theorem: rank + nullity = n

Matrix Inverse

• To solve Ax=b, we can write a closed-form solution if we can find a matrix A-1

such that AA-1 =A-1A=I (Identity matrix)• Then Ax=b iff x=A-1b:

x = Ix = A-1Ax = A-1b• A is non-singular iff A-1 exists iff Ax=b has a unique solution.• Note: If A-1,B-1 exist, then (AB)-1 = B-1A-1,

and (AT)-1 = (A-1)T

Note:

• For orthonormal matrices

A-1 = AT

Matrix Determinants

Note:• If det(A) = 0, then A is singular.• If det(A) ≠ 0, then A is invertible.• To compute:

– Simple example:

– Matlab: det(A)

bcaddcba

-=÷÷ø

öççè

ædet

MATLAB Interlude

Special Variables:

• Special variables:– ans : default variable name for the result– pi: p = 3.1415926…………– eps: Î = 2.2204e-016, smallest amount by which 2 numbers can differ.– Inf or inf : ¥, infinity– NaN or nan: not-a-number

Vectors: Example:>> x = [ 0 0.25*pi 0.5*pi 0.75*pi pi ]x =

0 0.7854 1.5708 2.3562 3.1416>> y = [ 0; 0.25*pi; 0.5*pi; 0.75*pi; pi ]y =

00.78541.57082.35623.1416

x is a row vector.

y is a column vector.

Vectors

• Vector Addressing – A vector element is addressed in MATLAB with an integer index enclosed in parentheses.

• Example:>> x(3)ans =

1.5708

ç 1st to 3rd elements of vector x

• The colon notation may be used to address a block of elements.(start : increment : end)

start is the starting index, increment is the amount to add to each successive index, and end is the ending index. A shortened format (start : end) may be used if increment is 1.

• Example:>> x(1:3)ans =

0 0.7854 1.5708

NOTE: MATLAB index starts at 1.

ç 3rd element of vector x

Vectors

Some useful commands:

x = start:end create row vector x starting with start, counting by one, ending at end

x = start:increment:end create row vector x starting with start, counting by increment, ending at or before end

linspace(start,end,number) create row vector x starting with start, ending at end, having number elements

length(x) returns the length of vector x

y = x’ transpose of vector x

dot (x, y) returns the scalar dot product of the vector x and y.

Matrices§ A Matrix array is two-dimensional, having both multiple rows and multiple columns,

similar to vector arrays:§ it begins with [, and end with ]§ spaces or commas are used to separate elements in a row§ semicolon or enter is used to separate rows.

•Example:•>> f = [ 1 2 3; 4 5 6]

f =1 2 34 5 6

• Matrix Addressing:-- matrixname(row, column)-- colon may be used in place of a row or column

reference to select the entire row or column.

recall:f =

1 2 34 5 6

h =2 4 61 3 5

n Example:

>> f(2,3)

ans =

6

>> h(:,1)

ans =

2

1

Matrices

Transpose B = A’Identity Matrix eye(n) è returns an n x n identity matrix

eye(m,n) è returns an m x n matrix with ones on the main diagonal and zeros elsewhere.

Addition and subtraction C = A + BC = A – B

Scalar Multiplication B = aA, where a is a scalar.Matrix Multiplication C = A*BMatrix Inverse B = inv(A), A must be a square matrix in this case.

rank (A) è returns the rank of the matrix A.Matrix Powers B = A.^2 è squares each element in the matrix

C = A * A è computes A*A, and A must be a square matrix.Determinant det (A), and A must be a square matrix.

more commands

A, B, C are matrices, and m, n, a are scalars.

Matrices

>> A = [1 2;3 4]A =

1 2 3 4

>> B = [2 0;2 1]B =

2 0 2 1

>> A*Bans =

6 2 14 4

>> A.*Bans =

2 0 6 4

>> A = [1 2 3;0 2 0]A = ��

��

>> B = [1;-1;0]B = ��

>> A*Bans = ��

Eigenvalues and Eigenvectors

• How can we characterize matrices?• The solutions to Ax = λx in the form of eigenpairs (λ,x) = (eigenvalue,eigenvector)

where x is non-zero• To solve this, (A – λI)x = 0• λ is an eigenvalue iff det(A – λI) = 0

Note:

Example:

÷÷÷

ø

ö

ççç

è

æ=

2/10064/30541

A

)2/1)(4/3)(1(2/10064/30541

)det( llll

ll

l ---=÷÷÷

ø

ö

ççç

è

æ

--

-=- IA

2/1,4/3,1 === lll

Eigenvalues and Eigenvectors

Example:

÷÷ø

öççè

æ=

1002

A Eigenvalues λ = 2, 1 with eigenvectors (1,0), (0,1)

Eigenvectors of a linear transformation A are not rotated (but will be scaled by the corresponding eigenvalue) when A is applied.

(0,1)

(2,0)

v Av

(1,0)

Important Results:

• Eigenvalues of nxn symmetric matrix are real but may not be distinct. There are n orthogonal eigenvectors.

Eigenvalues and EigenvectorsImportant Results:

• Eigenvalues of nxn symmetric matrix are real but may not be distinct. There are n orthogonal eigenvectors; e.g., Identity matrix

• Eigenvalues of a positive definite matrix are positive.

• MATLAB Command: [V,D] = eig(A)

• Suppose A has n linearly independent eigenvectors. Then

Q = [v1 v2 · · · vn]

) AQ = [�1v1 �2v2 · · · �nvn]

) AQ = Q⇤

) A = Q⇤Q-1

) Q-1AQ = ⇤

Note:

• The matrix Q thus provides a new basis for A.

Eigenvalues and EigenvectorsExample:>> A = [2 1 3;-2 0 -2;5 5 2]A = ��

��

>> X = A'*AX = ��

��

��Q,D] = eig(X)Q��

D = ��0

��

>>Q*D*inv(Q)ans =

��

>> Q’ans =

��

>> inv(Q)ans =

��

Eigenvalues and EigenvectorsCholesky Decomposition: Matrix must be positive definite>> A = [2 1 3;-2 0 -2;5 5 2]A = ��

��

>> X = A'*AX = ��

��

>> R = chol(X)R =��

>> R'*Rans =��

Solving Ax = b

Notes:

• MATLAB: x = A\b;• How stable is the solution?• If A or b are changed slightly, how much does it effect x?• The condition number c of A measures this:

c = λmax/ λmin

• Values of c near 1 are good.

>> A = [1 0;0 1e-10]A = ��

>> cond(A)ans = ��e+10>>

MATLAB

• Example: a system of 3 linear equations with 3 unknowns (x1, x2, x3):3x1 + 2x2 – x3 = 10-x1 + 3x2 + 2x3 = 5x1 – x2 – x3 = -1

Then, the system can be described as:

Ax = b

úúú

û

ù

êêê

ë

é

---=

111231123

A

úúú

û

ù

êêê

ë

é=

3

2

1

xxx

xúúú

û

ù

êêê

ë

é

-=

1510

b

Let :

MATLAB

• Solution by Matrix Inverse:Ax = bA-1Ax = A-1bx = A-1b

• MATLAB:>> A = [ 3 2 -1; -1 3 2; 1 -1 -1];>> b = [ 10; 5; -1];>> x = inv(A)*bx =

-2.00005.0000-6.0000

Answer:x1 = -2, x2 = 5, x3 = -6

• Solution by Matrix Division:The solution to the equation

Ax = bcan be computed using left division.

Answer:x1 = -2, x2 = 5, x3 = -6

n MATLAB:>> A = [ 3 2 -1; -1 3 2; 1 -1 -1];>> b = [ 10; 5; -1];>> x = A\bx =

-2.00005.0000-6.0000

MATLAB: Flow Control

for j=1:5 % use for-loops to execute iterations / repetitions

for i=1:3a(i, j) = i + j ;

end

end

For Loops:

If Conditional:a = zeros(3); b = zeros(3);

for j=1:3 for i=1:3

a(i,j) = rand; % use rand to generate a random number

if a(i,j) > 0.5 b(i,j) = 1;

end end

end

MATLABCell Arrays:

A cell array is a special array of arrays. Each element of the cell array may point to a scalar, an array, or another cell array.>> C = cell(2, 3); % create 2x3 empty cell array >> M = magic(2);>> a = 1:3; b = [4;5;6]; s = 'This is a string.';

>> C{1,1} = M; C{1,2} = a; C{2,1} = b; C{2,2} = s; C{1,3} = {1};C =

[2x2 double] [1x3 double] {1x1 cell}[2x1 double] ‘This is a string.‘ []

>> C{1,1} % prints contents of a specific cell element

ans =1 3

4 2>> C(1,:) % prints first row of cell array C; not its content

MATLABStructures:

Ideal layout for grouping arrays that are related.

>> name(1).last = ‘Smith’; name(2).last = ‘Hess’;>> name(1).first = ‘Mary’; name(2).first = ‘Robert’;>> name(1).sex = ‘female’; name(2).sex = ‘male’;

>> name(1).age = 45; name(2).age = 50;>> name(2)

ans = last: 'Hess'

first: 'Robert'

sex: 'male'age: 50

MATLABFrequently Used Functions:

>> magic(n) % creates a special n x n matrix; handy for testing

>> zeros(n,m) % creates n x m matrix of zeroes (0)>> ones(n,m) % creates n x m matrix of ones (1)>> rand(n,m) % creates n x m matrix of random numbers

>> repmat(a,n,m) % replicates a by n rows and m columns>> diag(M) % extracts the diagonals of a matrix M

>> help elmat % list all elementary matrix operations ( or elfun)>> abs(x); % absolute value of x >> exp(x); % e to the x-th power

>> fix(x); % rounds x to integer towards 0 >> log10(x); % common logarithm of x to the base 10

>> rem(x,y); % remainder of x/y>> mod(x, y); % modulus after division – unsigned rem>> sqrt(x); % square root of x

>> sin(x); % sine of x; x in radians >> acoth(x) % inversion hyperbolic cotangent of x

PlottingLine Plot:

>> t = 0:pi/100:2*pi;>> y = sin(t);>> plot(t,y)>> xlabel(‘t’);>> ylabel(‘sin(t)’);>> title(‘The plot of t vs sin(t)’);

PlottingCustomizing Graphical Effects

Generally, MATLAB’s default graphical settings are adequate which make plotting fairly effortless. For more customized effects, use the get and setcommands to change the behavior of specific rendering properties.>> hp1 = plot(1:5) % returns the handle of this line plot>> get(hp1) % to view line plot’s properties and their values>> set(hp1, ‘lineWidth’) % show possible values for lineWidth>> set(hp1, ‘lineWidth’, 2) % change line width of plot to 2>> gcf % returns current figure handle>> gca % returns current axes handle>> get(gcf) % gets current figure’s property settings>> set(gcf, ‘Name’, ‘My First Plot’) % Figure 1 => Figure 1: My First Plot>> get(gca) % gets the current axes’ property settings>> figure(1) % create/switch to Figure 1 or pop Figure 1 to the front >> clf % clears current figure>> close % close current figure; “close 3” closes Figure 3>> close all % close all figures

PlottingSurface Plot

>> Z = peaks; % generate data for plot; peaks returns function values>> surf(Z) % surface plot of Z

Try these commands also:>> shading flat>> shading interp>> shading faceted>> grid off>> axis off>> colorbar>> colormap(‘winter’)>> colormap(‘jet’)

PlottingContour Plots>> Z = peaks;>> contour(Z, 20) % contour plot of Z with 20 contours

>> contourf(Z, 20); % with color fill>> colormap('hot') % map option>> colorbar % make color bar

Singular Value Decomposition

TVUA S=

m´m m´n V is n´n

For an m´ n matrix A of rank r there exists a factorization(Singular Value Decomposition = SVD) as follows:

The columns of U are orthogonal eigenvectors of AAT.The columns of V are orthogonal eigenvectors of ATA.

ii ls =

( )rdiag ss ...1=S Singular values.

Eigenvalues l1 … lr of AAT are the eigenvalues of ATA.


n Illustration of SVD dimensions and sparseness


Let

úúú

û

ù

êêê

ë

é -=

011011

A

Thus m=3, n=2. Its SVD is

úû

ùêë

é

-úúú

û

ù

êêê

ë

é

úúú

û

ù

êêê

ë

é

--

2/12/12/12/1

003001

3/16/12/13/16/12/13/16/20

Typically, the singular values arranged in decreasing order.

MATLAB Linear AlgebraLinear Algebra and Numerical Matrix Theory Topics: Illustrate with MATLAB as...

Documents

Transcript of MATLAB Linear AlgebraLinear Algebra and Numerical Matrix Theory Topics: Illustrate with MATLAB as...